WP2 – Assessment and Psychological Mechanisms


1. To establish a knowledge base of psychological mechanisms contributing to PUI

2. To establish a gold-standard screening instrument for identifying adolescents at-risk for PUI

3. To identify subgroups of adolescents at-risk for PUI, who may benefit from semi- individualized intervention and prevention.

4. To calculate cost and burden of PUI.


WP2 consists of four main areas: Behavioural assessment and psychological mechanisms, digital phenotyping and mobile sensing, machine learning for risk profiling and assessment of cost and burden. Data from behavioural assessment and mobile sensing will be combined and analysed classically via testing specific hypotheses (in 2.1) and will be used in 2.3 to identify, using ML algorithms, subtypes of vulnerable individuals for whom specific interventions may work best, based on cognitive, affective, behavioural, and data tracking variables. The assessment of cost and burden of PUI will also integrate knowledge about at-risk and vulnerable groups based on WP 2.1-2.3. The overarching framework (Fig. 2) guides selection of the most important variables that will be assessed.

Task 2.1 Behavioural assessment & psy- chological mechanisms (M1-15)

Lead partner and participants UDE, UoS, MON, UoH, UZH, UPORTO, JGU Mainz, FIBHGM, ELTE, LSMU, VUA, CHUM

• Behavioural assessment includes a core battery of clinical assessment tools and cognitive tasks, conducted twice, at baseline (t1) and six-months follow-up (t2). In between (t1+3), a further assessment is conducted at which specific trait variables (e.g., personality, trait compulsivity) are examined combined with further moderating variables of interest (e.g., use expectancies, vaccine hesitancy). All scales/tasks are computer-/app- based and can be completed by participants unsupervised. In order not to burden the participants with too many questionnaires in a row, assessments at t1, t1+3, and t2 can be divided into 4 sessions of 15 minutes each.

• The core battery at t1 and t2 comprises scales related to PUI (PIUQ-SF-9, a short version of the ISAAQ, short CIUS), depression, anxiety, stress (DASS-21), cognitive tasks from BrainPAC (measuring inhibitory control, reward and reversal learning) and several other self-report scales assessing clinical variables of interest (e.g., emotional symptoms, attention deficits, obsessive-compulsive symptoms, cyberchondria, self-stigmatization). All scales are validated or considered valid for assessment of adolescents (for details see Annex 1).

• An ambulatory (end-of-day) assessment for 7 days will also be conducted after t1 and t2. The ambulatory assessment asks questions about the use of apps. over the course of the day, craving to use, related experiences of gratification and compensation, compulsive urges to use, mood and stress experiences, feelings of (unsuccessful) control over use. All items are rated on a 1-10 scale. Ambulatory assessment takes ca. 5 minutes.

• Data of the t1 and t2 core battery represent predictors and outcome variables (symptoms of PUI) in the analyses, while variables assessed at t1+3 will be used as mediating and moderating variables.

• Other variables of potential importance for understanding (risk of) PUI that cannot be captured by quantitative assessment are explored qualitatively in focus groups, such as effects of cosmetic issues (e.g., acne), interests and hobbies, adjustment to the pandemic, proneness for misinformation, internet literacy, family and friend’s constellations etc. WP2 will develop and pilot a guideline for conducting these focus groups (with WP4) and then disseminate the Guideline to all recruitment sites (WP1).

Task 2.2 Digital phenotyping and mobile sensing (M1-21)

Lead partner and participants UULM, IDC, SHEBA, UoH, UDE

• We will develop an application to record digital footprints of the study participants providing insights into a myriad of variables linked to PUI. Beyond actual app usage, this application will provide an overview on overall screen time, log in frequency, duration of each smartphone session and so forth over the 6 months behavioural assessment period.

• Beyond this, smartphone usage will be time stamped and the time variables can be also exploited in the later conducted analysis, for instance via ML in Task 2.3. From the recorded digital footprints not only insights into technology use can be provided, but also psychological states/traits can be “sensed” via the passive smartphone log data, as different forms of smartphone use have been associated with many important psychological variables.

• Of relevance for the overall project, not only will passive data be collected via the smartphone app. but also the administration of the batteries as described in Task 2.1 will be supported via smartphone and/or a web- application.

Participants can choose how to answer the questionnaire battery and can switch between modes.

Task 2.3 Machine learning (ML) (M1-M21)

Lead partner and participants IDC, UULM, SHEBA, UoH, UDE

• We will develop a ML-based algorithm to identify the at-risk vs. non-risk population and to tailor the most personalized intervention accordingly.

• We will apply ML to Cohort 1 to predict individuals at risk and identify actionable variables for application to subjects as intervention. We will develop multivariate predictive models using both classical (logistic regression, random forest) and modern methods (deep neural networks, domain adaptation, deep representations, adversarial training) to associate the assessments to the outcomes at the individual level. Generalization and validation will be assessed using common cross-validation methods at the patient level.

• The employment of ML will enable a massive data analysis, integrating both digital and clinical measurements in order to deliver algorithms that could (1) Identify the at-risk and not at-risk population to develop PUI and

(2) Identify the behavioural and psychological patterns at the basis of that risk in order to identify a personalized mechanism for tailored intervention.

• We will use the data of the behavioural assessment (PUI and clinical variables) at baseline (t1) and the 6- months follow-up (see description of task 1). In addition, to have more data points for PUI symptoms, the short CIUS (5 items) will be conducted monthly resulting in 7 measurements used for ML.

Task 2.4 Assessment of burden of PUI (M9 -21)

Lead partner and participants CAM, UDE, UoH

• We will assess the health burden of PUI by using validated health-related quality of life (HRQOL) scales and outcomes measures. We will be able to obtain from this the initial burden of PUI on participants to then see the possible beneficial effects of our intervention model (WP3). The two instruments will be used as direct patient reported outcomes of HRQOL and wellbeing at baseline (t1) and at 6 months (t2): the EQ-5D-Y which is a widely used instrument in cost utility analysis and health economic evaluation and the Paediatric Quality of Life Enjoyment and Satisfaction Questionnaire (PQ-LESQ), a measure of quality of life during the previous week.

Objectives and Ambition.

Our ambition is to mitigate the mental health burden of, and aid psychosocial adjustment to, the challenges of rapidly increasing digitalization among adolescents (aged 12-16y) as we look beyond the COVID-19 pandemic.

Bootstrap’s aim is to initiate health and social policy and practice change to reduce the harmful effects of digitalization on young people’s mental health.